Back to today’s briefing

Story perspectives

AI Risks Surge: 82% of Companies Face Data Breaches

8/27/2025

27 2

Summary

Story summary

Anthropic's AI model tests revealed risks, including potential blackmail from sensitive information access.
A Sailpoint survey indicated that 82% of companies using AI agents experienced unintended actions and data breaches.
Experts highlighted threats such as memory poisoning and tool misuse, stressing the importance of strong security measures.
OWASP identified 15 unique threats to agentic AI, underscoring increasing risks in AI deployment.

1 of 1

Story summary

Anthropic's AI model tests revealed risks, including potential blackmail from sensitive information access.
A Sailpoint survey indicated that 82% of companies using AI agents experienced unintended actions and data breaches.
Experts highlighted threats such as memory poisoning and tool misuse, stressing the importance of strong security measures.
OWASP identified 15 unique threats to agentic AI, underscoring increasing risks in AI deployment.